*** Sparse Interpretable Deep Learning with LIES Networks for Symbolic Regression ***

This repository contains the official implementation of the LIES Network for symbolic regression. LIES (Logarithm, Identity, Exponential, Sine) is a neural architecture designed to learn symbolic expressions by combining interpretable activations with sparsity-promoting training and pruning techniques.

🚀 Usage

First install the dependancies:

pip install -r requirements.txt

And then, run the pipeline with:

python lies_pipeline.py

🗃️ Dataset

We have tested LIES on the Feynman symbolic regression dataset found in this website.